Lang Hames [Wed, 3 Sep 2014 05:01:46 +0000 (05:01 +0000)]
[MCJIT] Add a 'section_addr' builtin function to RuntimeDyldChecker.
The syntax of the new builtin is 'section_addr(<filename>, <section-name>)'
(similar to the stub_addr builtin, but without a symbol name). It returns the
base address of the given section in the given object file. This builtin makes
it possible to refer to the contents of sections that cannot contain symbols,
e.g. sections added by the linker itself, like __eh_frame.
[FastISel][AArch64] Add target-dependent instruction selection for Add/Sub.
There is already target-dependent instruction selection support for Adds/Subs to
support compares and the intrinsics with overflow check. This takes advantage of
the existing infrastructure to also support Add/Sub, which allows the folding of
immediates, sign-/zero-extends, and shifts.
Hal Finkel [Tue, 2 Sep 2014 22:36:58 +0000 (22:36 +0000)]
[CFLAA] Remove tautological comparison
Fixes this (the warning is right, the unsigned value is not negative):
lib/Analysis/StratifiedSets.h:689:53: warning: comparison of unsigned expression >= 0 is always true [-Wtautological-compare]
bool inbounds(StratifiedIndex N) const { return N >= 0 && N < Links.size(); }
[FastISel][AArch64] Use the target-dependent selection code for shifts first.
This uses the target-dependent selection code for shifts first, which allows us
to create better code for shifts with immediates and sign-/zero-extend folding.
Vector type are not handled yet and the code falls back to target-independent
instruction selection for these cases.
Sean Silva [Tue, 2 Sep 2014 22:32:20 +0000 (22:32 +0000)]
Nuke MCAnalysis.
The code is buggy and barely tested. It is also mostly boilerplate.
(This includes MCObjectDisassembler, which is the interface to that
functionality)
Following an IRC discussion with Jim Grosbach, it seems sensible to just
nuke the whole lot of functionality, and dig it up from VCS if
necessary (I hope not!).
All of this stuff appears to have been added in a huge patch dump (look
at the timeframe surrounding e.g. r182628) where almost every patch
seemed to be untested and not reviewed before being committed.
Post-review responses to the patches were never addressed. I don't think
any of it would have passed pre-commit review.
I doubt anyone is depending on this, since this code appears to be
extremely buggy. In limited testing that Michael Spencer and I did, we
couldn't find a single real-world object file that wouldn't crash the
CFG reconstruction stuff. The symbolizer stuff has O(n^2) behavior and
so is not much use to anyone anyway. It seemed simpler to remove them as
a whole. Most of this code is boilerplate, which is the only way it was
able to scrape by 60% coverage.
HEADSUP: Modules folks, some files I nuked were referenced from
include/llvm/module.modulemap; I just deleted the references. Hopefully
that is the right fix (one was a FIXME though!).
Robin Morisset [Tue, 2 Sep 2014 22:16:29 +0000 (22:16 +0000)]
[X86] Allow atomic operations using immediates to avoid using a register
The only valid lowering of atomic stores in the X86 backend was mov from
register to memory. As a result, storing an immediate required a useless copy
of the immediate in a register. Now these can be compiled as a simple mov.
Similarily, adding/and-ing/or-ing/xor-ing an
immediate to an atomic location (but through an atomic_store/atomic_load,
not a fetch_whatever intrinsic) can now make use of an 'add $imm, x(%rip)'
instead of using a register. And the same applies to inc/dec.
This second point matches the first issue identified in
http://llvm.org/bugs/show_bug.cgi?id=17281
Hans Wennborg [Tue, 2 Sep 2014 21:51:35 +0000 (21:51 +0000)]
BumpPtrAllocator: use uintptr_t when aligning addresses to avoid undefined behaviour
In theory, alignPtr() could push a pointer beyond the end of the current slab, making
comparisons with that pointer undefined behaviour. Use an integer type to avoid this.
[asan] Assign a low branch weight to ASan's slow path, patch by Jonas Wagner. This speeds up asan (at least on SPEC) by 1%-5% or more. Also fix lint in dfsan.
Hal Finkel [Tue, 2 Sep 2014 21:43:13 +0000 (21:43 +0000)]
Add a CFL Alias Analysis implementation
This provides an implementation of CFL alias analysis (including some
supporting data structures). Currently, we don't have any extremely fancy
features, sans some interprocedural analysis (i.e. no field sensitivity, etc.),
and we do best sitting behind BasicAA + TBAA. In such a configuration, we take
~0.6-0.8% of total compile time, and give ~7-8% NoAlias responses to queries
TBAA and BasicAA couldn't answer when bootstrapping LLVM. In testing this on
other projects, we've seen up to 10.5% of queries dropped by BasicAA+TBAA
answered with NoAlias by this algorithm.
Patch by George Burgess IV (with minor modifications by me -- mostly adapting
some BasicAA tests), thanks!
[FastISel][AArch64] Move over to target-dependent instruction selection only.
This change moves FastISel for AArch64 to target-dependent instruction selection
only. This change replicates the existing target-independent behavior, therefore
there are no changes to the unit tests or new tests.
Future changes will take advantage of this change and update functionality
and unit tests.
[FastISel] Provide the option to skip target-independent instruction selection. NFC.
This allows the target to disable target-independent instruction selection and
jump directly into the target-dependent instruction selection code.
This can be beneficial for targets, such as AArch64, which could emit much
better code, but never got a chance to do so, because the target-independent
instruction selector was able to find an instruction sequence.
Refactor LowerFABS and LowerFNEG into one function (x86) (NFC)
We duplicate ~30 lines of code to lower FABS and FNEG for x86, so this patch combines them into one function.
No functional change intended, so no additional test cases. Test-suite behavior is unchanged.
Robin Morisset [Tue, 2 Sep 2014 20:17:52 +0000 (20:17 +0000)]
Fix MemoryDependenceAnalysis in cases where QueryInstr is a CmpXchg or a AtomicRMW
Summary:
MemoryDependenceAnalysis is currently cautious when the QueryInstr is an atomic
load or store, but I forgot to check for atomic cmpxchg/atomicrmw. This patch
is a way of fixing that, and making it less brittle (i.e. no risk that I forget
another possible kind of atomic, even if the IR ends up changing in the future),
by adding a fallback checking mayReadOrWriteFromMemory.
Thanks to Philip Reames for finding this bug and suggesting this solution in
http://reviews.llvm.org/D4845
Sadly, I don't see how to add a test for this, since the passes depending on
MemoryDependenceAnalysis won't trigger for an atomic rmw anyway. Does anyone
see a way for testing it?
"Setting" does not equal "copying". This bug has sat dormant for 2 reasons:
1. The unit test was not adequate.
2. Every current user of the "copyFastMathFlags" API is operating on a new instruction.
(ie, all existing fast-math flags are off). If you copy flags to an existing
instruction that has some flags on already, you will not necessarily turn them off
as expected.
I uncovered this bug while trying to implement a fix for PR20802.
Matt Arsenault [Tue, 2 Sep 2014 19:18:52 +0000 (19:18 +0000)]
Add note to documentation about machine node chains.
I've been assuming chain operands were always the first operand,
since the documentation says this. I was confused about why they
were missing after instruction selection. Apparently the convention
changes to using the last operand for MachineSDNodes and I've never
noticed before.
Matt Arsenault [Tue, 2 Sep 2014 19:02:53 +0000 (19:02 +0000)]
Fix interference caused by fmul 2, x -> fadd x, x
If an fmul was introduced by lowering, it wouldn't be folded
into a multiply by a constant since the earlier combine would
have replaced the fmul with the fadd.
Pete Cooper [Tue, 2 Sep 2014 17:43:54 +0000 (17:43 +0000)]
Change MCSchedModel to be a struct of statically initialized data.
This removes static initializers from the backends which generate this data, and also makes this struct match the other Tablegen generated structs in behaviour
David Blaikie [Tue, 2 Sep 2014 17:29:51 +0000 (17:29 +0000)]
Correct unique_ptr passing in MCObjectDisassembler::setFallbackRegion
Rather than passing by lvalue reference, pass by value to ensure that
the caller provides an rvalue (and use move assignment, rather than
release+reset, to assign to the member variable)
[APFloat] Fixed a bug in method 'fusedMultiplyAdd'.
When folding a fused multiply-add builtin call, make sure that we propagate the
correct result in the case where the addend is zero, and the two other operands
are finite non-zero.
Before this patch, the instruction simplifier wrongly folded the builtin call
in function @test to constant 'double 7.0'.
With this patch, method 'fusedMultiplyAdd' correctly evaluates the multiply and
propagates the expected result (i.e. 56.0).
Added test fold-builtin-fma.ll with the reproducible from PR20832 plus extra
test cases to verify the behavior of method 'fusedMultiplyAdd' in the presence
of NaN/Inf operands.
David Majnemer [Tue, 2 Sep 2014 16:22:00 +0000 (16:22 +0000)]
LICM: Don't crash when an instruction is used by an unreachable BB
Summary:
BBs might contain non-LCSSA'd values after the LCSSA pass is run if they
are unreachable from the entry block.
Normally, the users of the instruction would be PHIs but the unreachable
BBs have normal users; rewrite their uses to be undef values.
An alternative fix could involve fixing this at LCSSA but that would
require this invariant to hold after subsequent transforms. If a BB
created an unreachable block, they would be in violation of this.
Fix left shifts of negative integers in AArch64 InstPrinter/Disassembler
Summary:
Left shift of negative integer is an undefined behavior, and
is reported by UBSan. It's ok for imm values to be negative, so we can
just replace left shifts with multiplications.
Hal Finkel [Tue, 2 Sep 2014 16:05:23 +0000 (16:05 +0000)]
Enable splitting indexing from loads with TargetConstants
When I recommitted r208640 (in r216898) I added an exclusion for TargetConstant
offsets, as there is no guarantee that a backend can handle them on generic
ADDs (even if it generates them during address-mode matching) -- and,
specifically, applying this transformation directly with TargetConstants caused
a self-hosting failure on PPC64. Ignoring all TargetConstants, however, is less
than ideal. Instead, for non-opaque constants, we can convert them into regular
constants for use with the generated ADD (or SUB).
We have been using .init-array for most systems for quiet some time,
but tools like llc are still defaulting to .ctors because the old
option was never changed.
This patch makes llc default to .init-array and changes the option to
be -use-ctors.
Clang is not affected by this. It has its own fancier logic.
Hal Finkel [Tue, 2 Sep 2014 06:24:04 +0000 (06:24 +0000)]
Revert "Revert '[DAGCombiner] Split up an indexed load if only the base pointer value is live'"
I reverted r208640 in r209747 because r208640 broke self-hosting on PPC64. The
underlying cause of the failure is that pre-inc loads with increments
represented by ISD::TargetConstants were being transformed into ISD:::ADDs with
ISD::TargetConstant operands. PPC doesn't have a pattern for those, and so they
were selected as invalid r+r adds.
This recommits r208640, rebased and with an exclusion for ISD::TargetConstant
increments. This behavior seems correct, although in the future we might want
to ask the target to split out the indexing that uses ISD::TargetConstants.
Unfortunately, I don't yet have small test case where the relevant invalid
'add' instruction is not itself dead (and thus eliminated by
DeadMachineInstructionElim -- sometimes bugpoint is too good at removing things)
Original commit message (by Adam Nemet):
Right now the load may not get DCE'd because of the side-effect of updating
the base pointer.
This can happen if we lower a read-modify-write of an illegal larger type
(e.g. i48) such that the modification only affects one of the subparts (the
lower i32 part but not the higher i16 part). See the testcase.
In order to spot the dead load we need to revisit it when SimplifyDemandedBits
decided that the value of the load is masked off. This is the
CommitTargetLoweringOpt piece.
I checked compile time with ARM64 by sending SPEC bitcode files through llc.
No measurable change.
Hal Finkel [Tue, 2 Sep 2014 06:23:54 +0000 (06:23 +0000)]
[PowerPC] Guard against illegal selection of add for TargetConstant operands
r208640 was reverted because it caused a self-hosting failure on ppc64. The
underlying cause was the formation of ISD::ADD nodes with ISD::TargetConstant
operands. Because we have no patterns for 'add' taking 'timm' nodes, these are
selected as r+r add instructions (which is a miscompile). Guard against this
kind of behavior in the future by making the backend crash should this occur
(instead of silently generating invalid output).
The structures for Windows unwinding are shared across multiple platforms.
Indicate the encoding to be used for the particular target. Use this to switch
the unwind emitter instantiated by the AsmPrinter.
This is an enum class, and will be appropriately prefixed, making the encoding
type prefix redundant. No change to any uses as the use of this was not yet
introduced.
David Majnemer [Mon, 1 Sep 2014 21:10:02 +0000 (21:10 +0000)]
Revert "Revert two GEP-related InstCombine commits"
This reverts commit r216698 which reverted r216523 and r216598.
We would attempt to perform the transformation even if the match()
failed because, as a side effect, it would set V. This would trick us
into believing that we correctly found a place to correctly apply the
transform.
An additional test case was added to getelementptr.ll so that we might
not regress in the future.
This change will ease refactoring LowerFABS() and LowerFNEG()
since they have a lot of overlap.
Remove the creation of a floating point constant from an integer
because it's going to be used for a bitwise integer op anyway.
No change to codegen expected, but the verbose comment string
for asm output may change from float values to hex (integer),
depending on whether the constant already exists or not.
Add a convenience method to copy wrapping, exact, and fast-math flags (NFC).
The loop vectorizer preserves wrapping, exact, and fast-math properties of scalar instructions.
This patch adds a convenience method to make that operation easier because we need to do this
in the loop vectorizer, SLP vectorizer, and possibly other places.
Although this is a 'no functional change' patch, I've added a testcase to verify that the exact
flag is preserved by the loop vectorizer. The wrapping and fast-math flags are already checked
in existing testcases.
This patch implements a few changes related to the Thumb2 M-class MSR instruction:
* better handling of unpredictable encodings,
* recognition of the _g and _nzcvqg variants by the asm parser only if the DSP
extension is available, preferred output of MSR APSR moves with the _<bits>
suffix for v7-M.
Fix in InlineSpiller to make the rematerilization loop also consider
implicit uses of the whole register when a sub register is defined.
Now the same iterator is used in the rematerilization loop as in the
spill loop later.
Patch provided by Mikael Holmen.
This fix was proposed and reviewed by Quentin Colombet,
http://lists.cs.uiuc.edu/pipermail/llvmdev/2014-August/076135.html.
Unfortunately, this error in the rematerilization code has only been
seen in a large test case for an out-of-tree target, and is probably
hard to reproduce on an in-tree target. Therefore, no testcase is
provided.
Fix a really bad miscompile introduced in r216865 - the else-if logic
chain became completely broken here as *all* intrinsic users ended up
being skipped, and the ones that seemed to be singled out were actually
the exact wrong set.
This is a great example of why long else-if chains can be easily
confusing. Switch the entire code to use early exits and early continues
to have simpler (and more importantly, correct) logic here, as well as
fixing the reversed logic for detecting and continuing on lifetime
intrinsics.
I've also significantly cleaned up the test case and added another test
case demonstrating an example where the optimization is not (trivially)
safe to perform.
Small refactor on VectorizerHint for deduplication
Previously, the hint mechanism relied on clean up passes to remove redundant
metadata, which still showed up if running opt at low levels of optimization.
That also has shown that multiple nodes of the same type, but with different
values could still coexist, even if temporary, and cause confusion if the
next pass got the wrong value.
This patch makes sure that, if metadata already exists in a loop, the hint
mechanism will never append a new node, but always replace the existing one.
It also enhances the algorithm to cope with more metadata types in the future
by just adding a new type, not a lot of code.
Re-applying again due to MSVC 2013 being minimum requirement, and this patch
having C++11 that MSVC 2012 didn't support.
Hal Finkel [Mon, 1 Sep 2014 09:01:39 +0000 (09:01 +0000)]
Feed AA to the inliner and use AA->getModRefBehavior in AddAliasScopeMetadata
This feeds AA through the IFI structure into the inliner so that
AddAliasScopeMetadata can use AA->getModRefBehavior to figure out which
functions only access their arguments (instead of just hard-coding some
knowledge of memory intrinsics). Most of the information is only available from
BasicAA; this is important for preserving alias scoping information for
target-specific intrinsics when doing the noalias parameter attribute to
metadata conversion.
Nick Lewycky [Mon, 1 Sep 2014 05:17:15 +0000 (05:17 +0000)]
Remove an errant outer loop that contains nothing but an inner loop over exactly the same elements. While no functionality is change intended (and hence there are no changes to tests), you don't want to skip this revision if bisecting for errors.
Hal Finkel [Mon, 1 Sep 2014 04:26:40 +0000 (04:26 +0000)]
Fix AddAliasScopeMetadata again - alias.scope must be a complete description
I thought that I had fixed this problem in r216818, but I did not do a very
good job. The underlying issue is that when we add alias.scope metadata we are
asserting that this metadata completely describes the aliasing relationships
within the current aliasing scope domain, and so in the context of translating
noalias argument attributes, the pointers must all be based on noalias
arguments (as underlying objects) and have no other kind of underlying object.
In r216818 excluding appropriate accesses from getting alias.scope metadata is
done by looking for underlying objects that are not identified function-local
objects -- but that's wrong because allocas, etc. are also function-local
objects and we need to explicitly check that all underlying objects are the
noalias arguments for which we're adding metadata aliasing scopes.
This fixes the underlying-object check for adding alias.scope metadata, and
does some refactoring of the related capture-checking eligibility logic (and
adds more comments; hopefully making everything a bit clearer).
Fixes self-hosting on x86_64 with -mllvm -enable-noalias-to-md-conversion (the
feature is still disabled by default).
Summary:
Fixes a FIXME in MachineSinking. Instead of using the simple heuristics
in isPostDominatedBy, use the real MachinePostDominatorTree. The old
heuristics caused instructions to sink unnecessarily, and might create
register pressure.
Test Plan:
Added a NVPTX codegen test to verify that our change is in effect. It also
shows the unnecessary register pressure caused by over-sinking. Updated
affected tests in AArch64 and X86.
David Blaikie [Sun, 31 Aug 2014 21:26:22 +0000 (21:26 +0000)]
DebugInfo: Elide lexical scopes which only contain other (inline or lexical) scopes.
DW_TAG_lexical_scopes inform debuggers about the instruction range for
which a given variable (or imported declaration/module/etc) is valid. If
the scope doesn't itself contain any such entities, it's a waste of
space and should be omitted.
We were correctly doing this for entirely empty leaves, but not for
intermediate nodes.
Reduces total (not just debug sections) .o file size for a bootstrap
-gmlt LLVM by 22% and bootstrap -gmlt clang executable by 13%. The wins
for a full -g build will be less as a % (and in absolute terms), but
should still be substantial - with some of that win being fewer
relocations, thus more substantiall reducing link times than fewer bytes
alone would have.
David Blaikie [Sun, 31 Aug 2014 18:04:28 +0000 (18:04 +0000)]
DebugInfo: Move argument creation up into the caller that's unambiguously handling the subprogram scope (replacing a conditional with an assertion in the process)
David Blaikie [Sun, 31 Aug 2014 05:46:17 +0000 (05:46 +0000)]
Delay adding imported entity DIEs to the lexical scope, streamlining the check for "this scope has nothing in it"
This makes the emptiness of the scope with regards to variables and
nested scopes is the same as with regards to imported entities. Just
check if we had nothing at all before we build the node.
David Blaikie [Sun, 31 Aug 2014 05:32:06 +0000 (05:32 +0000)]
Refactor constructImportedEntityDIE(DwarfUnit, DIImportedEntity) to return a DIE rather than inserting it into a specified context.
First of many steps to improve lexical scope construction (to omit
trivial lexical scopes - those without any direct variables). To that
end it's easier not to create imported entities directly into the
lexical scope node, but to build them, then add them if necessary.
Jordan Rose [Sat, 30 Aug 2014 17:07:55 +0000 (17:07 +0000)]
Teach llvm-bcanalyzer to use one stream's BLOCKINFO to read another stream.
This allows streams that only use BLOCKINFO for debugging purposes to omit
the block entirely. As long as another stream is available with the correct
BLOCKINFO, the first stream can still be analyzed and dumped.
As part of this commit, BitstreamReader gets a move constructor and move
assignment operator, as well as a takeBlockInfo method.
projects / llvm / log